2 * Copyright (c) 2007 Mellanox Technologies. All rights reserved.
4 * This software is available to you under a choice of one of two
5 * licenses. You may choose to be licensed under the terms of the GNU
6 * General Public License (GPL) Version 2, available from the file
7 * COPYING in the main directory of this source tree, or the
8 * OpenIB.org BSD license below:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
18 * - Redistributions in binary form must reproduce the above
19 * copyright notice, this list of conditions and the following
20 * disclaimer in the documentation and/or other materials
21 * provided with the distribution.
23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
34 #include <linux/etherdevice.h>
35 #include <linux/tcp.h>
36 #include <linux/if_vlan.h>
37 #include <linux/delay.h>
38 #include <linux/slab.h>
39 #include <linux/hash.h>
41 #include <net/busy_poll.h>
42 #include <net/vxlan.h>
44 #include <linux/mlx4/driver.h>
45 #include <linux/mlx4/device.h>
46 #include <linux/mlx4/cmd.h>
47 #include <linux/mlx4/cq.h>
52 int mlx4_en_setup_tc(struct net_device *dev, u8 up)
54 struct mlx4_en_priv *priv = netdev_priv(dev);
56 unsigned int offset = 0;
58 if (up && up != MLX4_EN_NUM_UP)
61 netdev_set_num_tc(dev, up);
63 /* Partition Tx queues evenly amongst UP's */
64 for (i = 0; i < up; i++) {
65 netdev_set_tc_queue(dev, i, priv->num_tx_rings_p_up, offset);
66 offset += priv->num_tx_rings_p_up;
72 #ifdef CONFIG_NET_RX_BUSY_POLL
73 /* must be called with local_bh_disable()d */
74 static int mlx4_en_low_latency_recv(struct napi_struct *napi)
76 struct mlx4_en_cq *cq = container_of(napi, struct mlx4_en_cq, napi);
77 struct net_device *dev = cq->dev;
78 struct mlx4_en_priv *priv = netdev_priv(dev);
79 struct mlx4_en_rx_ring *rx_ring = priv->rx_ring[cq->ring];
83 return LL_FLUSH_FAILED;
85 if (!mlx4_en_cq_lock_poll(cq))
88 done = mlx4_en_process_rx_cq(dev, cq, 4);
90 rx_ring->cleaned += done;
94 mlx4_en_cq_unlock_poll(cq);
98 #endif /* CONFIG_NET_RX_BUSY_POLL */
100 #ifdef CONFIG_RFS_ACCEL
102 struct mlx4_en_filter {
103 struct list_head next;
104 struct work_struct work;
113 struct mlx4_en_priv *priv;
114 u32 flow_id; /* RFS infrastructure id */
115 int id; /* mlx4_en driver id */
116 u64 reg_id; /* Flow steering API id */
117 u8 activated; /* Used to prevent expiry before filter
120 struct hlist_node filter_chain;
123 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv);
125 static enum mlx4_net_trans_rule_id mlx4_ip_proto_to_trans_rule_id(u8 ip_proto)
129 return MLX4_NET_TRANS_RULE_ID_UDP;
131 return MLX4_NET_TRANS_RULE_ID_TCP;
133 return MLX4_NET_TRANS_RULE_NUM;
137 static void mlx4_en_filter_work(struct work_struct *work)
139 struct mlx4_en_filter *filter = container_of(work,
140 struct mlx4_en_filter,
142 struct mlx4_en_priv *priv = filter->priv;
143 struct mlx4_spec_list spec_tcp_udp = {
144 .id = mlx4_ip_proto_to_trans_rule_id(filter->ip_proto),
147 .dst_port = filter->dst_port,
148 .dst_port_msk = (__force __be16)-1,
149 .src_port = filter->src_port,
150 .src_port_msk = (__force __be16)-1,
154 struct mlx4_spec_list spec_ip = {
155 .id = MLX4_NET_TRANS_RULE_ID_IPV4,
158 .dst_ip = filter->dst_ip,
159 .dst_ip_msk = (__force __be32)-1,
160 .src_ip = filter->src_ip,
161 .src_ip_msk = (__force __be32)-1,
165 struct mlx4_spec_list spec_eth = {
166 .id = MLX4_NET_TRANS_RULE_ID_ETH,
168 struct mlx4_net_trans_rule rule = {
169 .list = LIST_HEAD_INIT(rule.list),
170 .queue_mode = MLX4_NET_TRANS_Q_LIFO,
173 .promisc_mode = MLX4_FS_REGULAR,
175 .priority = MLX4_DOMAIN_RFS,
178 __be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
180 if (spec_tcp_udp.id >= MLX4_NET_TRANS_RULE_NUM) {
181 en_warn(priv, "RFS: ignoring unsupported ip protocol (%d)\n",
185 list_add_tail(&spec_eth.list, &rule.list);
186 list_add_tail(&spec_ip.list, &rule.list);
187 list_add_tail(&spec_tcp_udp.list, &rule.list);
189 rule.qpn = priv->rss_map.qps[filter->rxq_index].qpn;
190 memcpy(spec_eth.eth.dst_mac, priv->dev->dev_addr, ETH_ALEN);
191 memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
193 filter->activated = 0;
195 if (filter->reg_id) {
196 rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
197 if (rc && rc != -ENOENT)
198 en_err(priv, "Error detaching flow. rc = %d\n", rc);
201 rc = mlx4_flow_attach(priv->mdev->dev, &rule, &filter->reg_id);
203 en_err(priv, "Error attaching flow. err = %d\n", rc);
206 mlx4_en_filter_rfs_expire(priv);
208 filter->activated = 1;
211 static inline struct hlist_head *
212 filter_hash_bucket(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
213 __be16 src_port, __be16 dst_port)
218 l = (__force unsigned long)src_port |
219 ((__force unsigned long)dst_port << 2);
220 l ^= (__force unsigned long)(src_ip ^ dst_ip);
222 bucket_idx = hash_long(l, MLX4_EN_FILTER_HASH_SHIFT);
224 return &priv->filter_hash[bucket_idx];
227 static struct mlx4_en_filter *
228 mlx4_en_filter_alloc(struct mlx4_en_priv *priv, int rxq_index, __be32 src_ip,
229 __be32 dst_ip, u8 ip_proto, __be16 src_port,
230 __be16 dst_port, u32 flow_id)
232 struct mlx4_en_filter *filter = NULL;
234 filter = kzalloc(sizeof(struct mlx4_en_filter), GFP_ATOMIC);
239 filter->rxq_index = rxq_index;
240 INIT_WORK(&filter->work, mlx4_en_filter_work);
242 filter->src_ip = src_ip;
243 filter->dst_ip = dst_ip;
244 filter->ip_proto = ip_proto;
245 filter->src_port = src_port;
246 filter->dst_port = dst_port;
248 filter->flow_id = flow_id;
250 filter->id = priv->last_filter_id++ % RPS_NO_FILTER;
252 list_add_tail(&filter->next, &priv->filters);
253 hlist_add_head(&filter->filter_chain,
254 filter_hash_bucket(priv, src_ip, dst_ip, src_port,
260 static void mlx4_en_filter_free(struct mlx4_en_filter *filter)
262 struct mlx4_en_priv *priv = filter->priv;
265 list_del(&filter->next);
267 rc = mlx4_flow_detach(priv->mdev->dev, filter->reg_id);
268 if (rc && rc != -ENOENT)
269 en_err(priv, "Error detaching flow. rc = %d\n", rc);
274 static inline struct mlx4_en_filter *
275 mlx4_en_filter_find(struct mlx4_en_priv *priv, __be32 src_ip, __be32 dst_ip,
276 u8 ip_proto, __be16 src_port, __be16 dst_port)
278 struct mlx4_en_filter *filter;
279 struct mlx4_en_filter *ret = NULL;
281 hlist_for_each_entry(filter,
282 filter_hash_bucket(priv, src_ip, dst_ip,
285 if (filter->src_ip == src_ip &&
286 filter->dst_ip == dst_ip &&
287 filter->ip_proto == ip_proto &&
288 filter->src_port == src_port &&
289 filter->dst_port == dst_port) {
299 mlx4_en_filter_rfs(struct net_device *net_dev, const struct sk_buff *skb,
300 u16 rxq_index, u32 flow_id)
302 struct mlx4_en_priv *priv = netdev_priv(net_dev);
303 struct mlx4_en_filter *filter;
304 const struct iphdr *ip;
311 int nhoff = skb_network_offset(skb);
314 if (skb->protocol != htons(ETH_P_IP))
315 return -EPROTONOSUPPORT;
317 ip = (const struct iphdr *)(skb->data + nhoff);
318 if (ip_is_fragment(ip))
319 return -EPROTONOSUPPORT;
321 if ((ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
322 return -EPROTONOSUPPORT;
323 ports = (const __be16 *)(skb->data + nhoff + 4 * ip->ihl);
325 ip_proto = ip->protocol;
331 spin_lock_bh(&priv->filters_lock);
332 filter = mlx4_en_filter_find(priv, src_ip, dst_ip, ip_proto,
335 if (filter->rxq_index == rxq_index)
338 filter->rxq_index = rxq_index;
340 filter = mlx4_en_filter_alloc(priv, rxq_index,
341 src_ip, dst_ip, ip_proto,
342 src_port, dst_port, flow_id);
349 queue_work(priv->mdev->workqueue, &filter->work);
354 spin_unlock_bh(&priv->filters_lock);
359 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv)
361 struct mlx4_en_filter *filter, *tmp;
364 spin_lock_bh(&priv->filters_lock);
365 list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
366 list_move(&filter->next, &del_list);
367 hlist_del(&filter->filter_chain);
369 spin_unlock_bh(&priv->filters_lock);
371 list_for_each_entry_safe(filter, tmp, &del_list, next) {
372 cancel_work_sync(&filter->work);
373 mlx4_en_filter_free(filter);
377 static void mlx4_en_filter_rfs_expire(struct mlx4_en_priv *priv)
379 struct mlx4_en_filter *filter = NULL, *tmp, *last_filter = NULL;
383 spin_lock_bh(&priv->filters_lock);
384 list_for_each_entry_safe(filter, tmp, &priv->filters, next) {
385 if (i > MLX4_EN_FILTER_EXPIRY_QUOTA)
388 if (filter->activated &&
389 !work_pending(&filter->work) &&
390 rps_may_expire_flow(priv->dev,
391 filter->rxq_index, filter->flow_id,
393 list_move(&filter->next, &del_list);
394 hlist_del(&filter->filter_chain);
396 last_filter = filter;
401 if (last_filter && (&last_filter->next != priv->filters.next))
402 list_move(&priv->filters, &last_filter->next);
404 spin_unlock_bh(&priv->filters_lock);
406 list_for_each_entry_safe(filter, tmp, &del_list, next)
407 mlx4_en_filter_free(filter);
411 static int mlx4_en_vlan_rx_add_vid(struct net_device *dev,
412 __be16 proto, u16 vid)
414 struct mlx4_en_priv *priv = netdev_priv(dev);
415 struct mlx4_en_dev *mdev = priv->mdev;
419 en_dbg(HW, priv, "adding VLAN:%d\n", vid);
421 set_bit(vid, priv->active_vlans);
423 /* Add VID to port VLAN filter */
424 mutex_lock(&mdev->state_lock);
425 if (mdev->device_up && priv->port_up) {
426 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
428 en_err(priv, "Failed configuring VLAN filter\n");
430 if (mlx4_register_vlan(mdev->dev, priv->port, vid, &idx))
431 en_dbg(HW, priv, "failed adding vlan %d\n", vid);
432 mutex_unlock(&mdev->state_lock);
437 static int mlx4_en_vlan_rx_kill_vid(struct net_device *dev,
438 __be16 proto, u16 vid)
440 struct mlx4_en_priv *priv = netdev_priv(dev);
441 struct mlx4_en_dev *mdev = priv->mdev;
444 en_dbg(HW, priv, "Killing VID:%d\n", vid);
446 clear_bit(vid, priv->active_vlans);
448 /* Remove VID from port VLAN filter */
449 mutex_lock(&mdev->state_lock);
450 mlx4_unregister_vlan(mdev->dev, priv->port, vid);
452 if (mdev->device_up && priv->port_up) {
453 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
455 en_err(priv, "Failed configuring VLAN filter\n");
457 mutex_unlock(&mdev->state_lock);
462 static void mlx4_en_u64_to_mac(unsigned char dst_mac[ETH_ALEN + 2], u64 src_mac)
465 for (i = ETH_ALEN - 1; i >= 0; --i) {
466 dst_mac[i] = src_mac & 0xff;
469 memset(&dst_mac[ETH_ALEN], 0, 2);
473 static int mlx4_en_tunnel_steer_add(struct mlx4_en_priv *priv, unsigned char *addr,
474 int qpn, u64 *reg_id)
477 struct mlx4_spec_list spec_eth_outer = { {NULL} };
478 struct mlx4_spec_list spec_vxlan = { {NULL} };
479 struct mlx4_spec_list spec_eth_inner = { {NULL} };
481 struct mlx4_net_trans_rule rule = {
482 .queue_mode = MLX4_NET_TRANS_Q_FIFO,
485 .promisc_mode = MLX4_FS_REGULAR,
486 .priority = MLX4_DOMAIN_NIC,
489 __be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
491 if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
492 return 0; /* do nothing */
494 rule.port = priv->port;
496 INIT_LIST_HEAD(&rule.list);
498 spec_eth_outer.id = MLX4_NET_TRANS_RULE_ID_ETH;
499 memcpy(spec_eth_outer.eth.dst_mac, addr, ETH_ALEN);
500 memcpy(spec_eth_outer.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
502 spec_vxlan.id = MLX4_NET_TRANS_RULE_ID_VXLAN; /* any vxlan header */
503 spec_eth_inner.id = MLX4_NET_TRANS_RULE_ID_ETH; /* any inner eth header */
505 list_add_tail(&spec_eth_outer.list, &rule.list);
506 list_add_tail(&spec_vxlan.list, &rule.list);
507 list_add_tail(&spec_eth_inner.list, &rule.list);
509 err = mlx4_flow_attach(priv->mdev->dev, &rule, reg_id);
511 en_err(priv, "failed to add vxlan steering rule, err %d\n", err);
514 en_dbg(DRV, priv, "added vxlan steering rule, mac %pM reg_id %llx\n", addr, *reg_id);
519 static int mlx4_en_uc_steer_add(struct mlx4_en_priv *priv,
520 unsigned char *mac, int *qpn, u64 *reg_id)
522 struct mlx4_en_dev *mdev = priv->mdev;
523 struct mlx4_dev *dev = mdev->dev;
526 switch (dev->caps.steering_mode) {
527 case MLX4_STEERING_MODE_B0: {
532 memcpy(&gid[10], mac, ETH_ALEN);
535 err = mlx4_unicast_attach(dev, &qp, gid, 0, MLX4_PROT_ETH);
538 case MLX4_STEERING_MODE_DEVICE_MANAGED: {
539 struct mlx4_spec_list spec_eth = { {NULL} };
540 __be64 mac_mask = cpu_to_be64(MLX4_MAC_MASK << 16);
542 struct mlx4_net_trans_rule rule = {
543 .queue_mode = MLX4_NET_TRANS_Q_FIFO,
546 .promisc_mode = MLX4_FS_REGULAR,
547 .priority = MLX4_DOMAIN_NIC,
550 rule.port = priv->port;
552 INIT_LIST_HEAD(&rule.list);
554 spec_eth.id = MLX4_NET_TRANS_RULE_ID_ETH;
555 memcpy(spec_eth.eth.dst_mac, mac, ETH_ALEN);
556 memcpy(spec_eth.eth.dst_mac_msk, &mac_mask, ETH_ALEN);
557 list_add_tail(&spec_eth.list, &rule.list);
559 err = mlx4_flow_attach(dev, &rule, reg_id);
566 en_warn(priv, "Failed Attaching Unicast\n");
571 static void mlx4_en_uc_steer_release(struct mlx4_en_priv *priv,
572 unsigned char *mac, int qpn, u64 reg_id)
574 struct mlx4_en_dev *mdev = priv->mdev;
575 struct mlx4_dev *dev = mdev->dev;
577 switch (dev->caps.steering_mode) {
578 case MLX4_STEERING_MODE_B0: {
583 memcpy(&gid[10], mac, ETH_ALEN);
586 mlx4_unicast_detach(dev, &qp, gid, MLX4_PROT_ETH);
589 case MLX4_STEERING_MODE_DEVICE_MANAGED: {
590 mlx4_flow_detach(dev, reg_id);
594 en_err(priv, "Invalid steering mode.\n");
598 static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
600 struct mlx4_en_dev *mdev = priv->mdev;
601 struct mlx4_dev *dev = mdev->dev;
602 struct mlx4_mac_entry *entry;
606 int *qpn = &priv->base_qpn;
607 u64 mac = mlx4_mac_to_u64(priv->dev->dev_addr);
609 en_dbg(DRV, priv, "Registering MAC: %pM for adding\n",
610 priv->dev->dev_addr);
611 index = mlx4_register_mac(dev, priv->port, mac);
614 en_err(priv, "Failed adding MAC: %pM\n",
615 priv->dev->dev_addr);
619 if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
620 int base_qpn = mlx4_get_base_qpn(dev, priv->port);
621 *qpn = base_qpn + index;
625 err = mlx4_qp_reserve_range(dev, 1, 1, qpn);
626 en_dbg(DRV, priv, "Reserved qp %d\n", *qpn);
628 en_err(priv, "Failed to reserve qp for mac registration\n");
632 err = mlx4_en_uc_steer_add(priv, priv->dev->dev_addr, qpn, ®_id);
636 err = mlx4_en_tunnel_steer_add(priv, priv->dev->dev_addr, *qpn,
637 &priv->tunnel_reg_id);
641 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
646 memcpy(entry->mac, priv->dev->dev_addr, sizeof(entry->mac));
647 entry->reg_id = reg_id;
649 hlist_add_head_rcu(&entry->hlist,
650 &priv->mac_hash[entry->mac[MLX4_EN_MAC_HASH_IDX]]);
655 if (priv->tunnel_reg_id)
656 mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
658 mlx4_en_uc_steer_release(priv, priv->dev->dev_addr, *qpn, reg_id);
661 mlx4_qp_release_range(dev, *qpn, 1);
664 mlx4_unregister_mac(dev, priv->port, mac);
668 static void mlx4_en_put_qp(struct mlx4_en_priv *priv)
670 struct mlx4_en_dev *mdev = priv->mdev;
671 struct mlx4_dev *dev = mdev->dev;
672 int qpn = priv->base_qpn;
675 if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
676 mac = mlx4_mac_to_u64(priv->dev->dev_addr);
677 en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
678 priv->dev->dev_addr);
679 mlx4_unregister_mac(dev, priv->port, mac);
681 struct mlx4_mac_entry *entry;
682 struct hlist_node *tmp;
683 struct hlist_head *bucket;
686 for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
687 bucket = &priv->mac_hash[i];
688 hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
689 mac = mlx4_mac_to_u64(entry->mac);
690 en_dbg(DRV, priv, "Registering MAC: %pM for deleting\n",
692 mlx4_en_uc_steer_release(priv, entry->mac,
695 mlx4_unregister_mac(dev, priv->port, mac);
696 hlist_del_rcu(&entry->hlist);
697 kfree_rcu(entry, rcu);
701 if (priv->tunnel_reg_id) {
702 mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
703 priv->tunnel_reg_id = 0;
706 en_dbg(DRV, priv, "Releasing qp: port %d, qpn %d\n",
708 mlx4_qp_release_range(dev, qpn, 1);
709 priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
713 static int mlx4_en_replace_mac(struct mlx4_en_priv *priv, int qpn,
714 unsigned char *new_mac, unsigned char *prev_mac)
716 struct mlx4_en_dev *mdev = priv->mdev;
717 struct mlx4_dev *dev = mdev->dev;
719 u64 new_mac_u64 = mlx4_mac_to_u64(new_mac);
721 if (dev->caps.steering_mode != MLX4_STEERING_MODE_A0) {
722 struct hlist_head *bucket;
723 unsigned int mac_hash;
724 struct mlx4_mac_entry *entry;
725 struct hlist_node *tmp;
726 u64 prev_mac_u64 = mlx4_mac_to_u64(prev_mac);
728 bucket = &priv->mac_hash[prev_mac[MLX4_EN_MAC_HASH_IDX]];
729 hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
730 if (ether_addr_equal_64bits(entry->mac, prev_mac)) {
731 mlx4_en_uc_steer_release(priv, entry->mac,
733 mlx4_unregister_mac(dev, priv->port,
735 hlist_del_rcu(&entry->hlist);
737 memcpy(entry->mac, new_mac, ETH_ALEN);
739 mac_hash = new_mac[MLX4_EN_MAC_HASH_IDX];
740 hlist_add_head_rcu(&entry->hlist,
741 &priv->mac_hash[mac_hash]);
742 mlx4_register_mac(dev, priv->port, new_mac_u64);
743 err = mlx4_en_uc_steer_add(priv, new_mac,
748 if (priv->tunnel_reg_id) {
749 mlx4_flow_detach(priv->mdev->dev, priv->tunnel_reg_id);
750 priv->tunnel_reg_id = 0;
752 err = mlx4_en_tunnel_steer_add(priv, new_mac, qpn,
753 &priv->tunnel_reg_id);
760 return __mlx4_replace_mac(dev, priv->port, qpn, new_mac_u64);
763 static int mlx4_en_do_set_mac(struct mlx4_en_priv *priv)
768 /* Remove old MAC and insert the new one */
769 err = mlx4_en_replace_mac(priv, priv->base_qpn,
770 priv->dev->dev_addr, priv->prev_mac);
772 en_err(priv, "Failed changing HW MAC address\n");
774 en_dbg(HW, priv, "Port is down while registering mac, exiting...\n");
776 memcpy(priv->prev_mac, priv->dev->dev_addr,
777 sizeof(priv->prev_mac));
782 static int mlx4_en_set_mac(struct net_device *dev, void *addr)
784 struct mlx4_en_priv *priv = netdev_priv(dev);
785 struct mlx4_en_dev *mdev = priv->mdev;
786 struct sockaddr *saddr = addr;
789 if (!is_valid_ether_addr(saddr->sa_data))
790 return -EADDRNOTAVAIL;
792 mutex_lock(&mdev->state_lock);
793 memcpy(dev->dev_addr, saddr->sa_data, ETH_ALEN);
794 err = mlx4_en_do_set_mac(priv);
795 mutex_unlock(&mdev->state_lock);
800 static void mlx4_en_clear_list(struct net_device *dev)
802 struct mlx4_en_priv *priv = netdev_priv(dev);
803 struct mlx4_en_mc_list *tmp, *mc_to_del;
805 list_for_each_entry_safe(mc_to_del, tmp, &priv->mc_list, list) {
806 list_del(&mc_to_del->list);
811 static void mlx4_en_cache_mclist(struct net_device *dev)
813 struct mlx4_en_priv *priv = netdev_priv(dev);
814 struct netdev_hw_addr *ha;
815 struct mlx4_en_mc_list *tmp;
817 mlx4_en_clear_list(dev);
818 netdev_for_each_mc_addr(ha, dev) {
819 tmp = kzalloc(sizeof(struct mlx4_en_mc_list), GFP_ATOMIC);
821 mlx4_en_clear_list(dev);
824 memcpy(tmp->addr, ha->addr, ETH_ALEN);
825 list_add_tail(&tmp->list, &priv->mc_list);
829 static void update_mclist_flags(struct mlx4_en_priv *priv,
830 struct list_head *dst,
831 struct list_head *src)
833 struct mlx4_en_mc_list *dst_tmp, *src_tmp, *new_mc;
836 /* Find all the entries that should be removed from dst,
837 * These are the entries that are not found in src
839 list_for_each_entry(dst_tmp, dst, list) {
841 list_for_each_entry(src_tmp, src, list) {
842 if (ether_addr_equal(dst_tmp->addr, src_tmp->addr)) {
848 dst_tmp->action = MCLIST_REM;
851 /* Add entries that exist in src but not in dst
852 * mark them as need to add
854 list_for_each_entry(src_tmp, src, list) {
856 list_for_each_entry(dst_tmp, dst, list) {
857 if (ether_addr_equal(dst_tmp->addr, src_tmp->addr)) {
858 dst_tmp->action = MCLIST_NONE;
864 new_mc = kmemdup(src_tmp,
865 sizeof(struct mlx4_en_mc_list),
870 new_mc->action = MCLIST_ADD;
871 list_add_tail(&new_mc->list, dst);
876 static void mlx4_en_set_rx_mode(struct net_device *dev)
878 struct mlx4_en_priv *priv = netdev_priv(dev);
883 queue_work(priv->mdev->workqueue, &priv->rx_mode_task);
886 static void mlx4_en_set_promisc_mode(struct mlx4_en_priv *priv,
887 struct mlx4_en_dev *mdev)
891 if (!(priv->flags & MLX4_EN_FLAG_PROMISC)) {
892 if (netif_msg_rx_status(priv))
893 en_warn(priv, "Entering promiscuous mode\n");
894 priv->flags |= MLX4_EN_FLAG_PROMISC;
896 /* Enable promiscouos mode */
897 switch (mdev->dev->caps.steering_mode) {
898 case MLX4_STEERING_MODE_DEVICE_MANAGED:
899 err = mlx4_flow_steer_promisc_add(mdev->dev,
902 MLX4_FS_ALL_DEFAULT);
904 en_err(priv, "Failed enabling promiscuous mode\n");
905 priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
908 case MLX4_STEERING_MODE_B0:
909 err = mlx4_unicast_promisc_add(mdev->dev,
913 en_err(priv, "Failed enabling unicast promiscuous mode\n");
915 /* Add the default qp number as multicast
918 if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
919 err = mlx4_multicast_promisc_add(mdev->dev,
923 en_err(priv, "Failed enabling multicast promiscuous mode\n");
924 priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
928 case MLX4_STEERING_MODE_A0:
929 err = mlx4_SET_PORT_qpn_calc(mdev->dev,
934 en_err(priv, "Failed enabling promiscuous mode\n");
938 /* Disable port multicast filter (unconditionally) */
939 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
940 0, MLX4_MCAST_DISABLE);
942 en_err(priv, "Failed disabling multicast filter\n");
944 /* Disable port VLAN filter */
945 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
947 en_err(priv, "Failed disabling VLAN filter\n");
951 static void mlx4_en_clear_promisc_mode(struct mlx4_en_priv *priv,
952 struct mlx4_en_dev *mdev)
956 if (netif_msg_rx_status(priv))
957 en_warn(priv, "Leaving promiscuous mode\n");
958 priv->flags &= ~MLX4_EN_FLAG_PROMISC;
960 /* Disable promiscouos mode */
961 switch (mdev->dev->caps.steering_mode) {
962 case MLX4_STEERING_MODE_DEVICE_MANAGED:
963 err = mlx4_flow_steer_promisc_remove(mdev->dev,
965 MLX4_FS_ALL_DEFAULT);
967 en_err(priv, "Failed disabling promiscuous mode\n");
968 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
971 case MLX4_STEERING_MODE_B0:
972 err = mlx4_unicast_promisc_remove(mdev->dev,
976 en_err(priv, "Failed disabling unicast promiscuous mode\n");
977 /* Disable Multicast promisc */
978 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
979 err = mlx4_multicast_promisc_remove(mdev->dev,
983 en_err(priv, "Failed disabling multicast promiscuous mode\n");
984 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
988 case MLX4_STEERING_MODE_A0:
989 err = mlx4_SET_PORT_qpn_calc(mdev->dev,
993 en_err(priv, "Failed disabling promiscuous mode\n");
997 /* Enable port VLAN filter */
998 err = mlx4_SET_VLAN_FLTR(mdev->dev, priv);
1000 en_err(priv, "Failed enabling VLAN filter\n");
1003 static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
1004 struct net_device *dev,
1005 struct mlx4_en_dev *mdev)
1007 struct mlx4_en_mc_list *mclist, *tmp;
1009 u8 mc_list[16] = {0};
1012 /* Enable/disable the multicast filter according to IFF_ALLMULTI */
1013 if (dev->flags & IFF_ALLMULTI) {
1014 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
1015 0, MLX4_MCAST_DISABLE);
1017 en_err(priv, "Failed disabling multicast filter\n");
1019 /* Add the default qp number as multicast promisc */
1020 if (!(priv->flags & MLX4_EN_FLAG_MC_PROMISC)) {
1021 switch (mdev->dev->caps.steering_mode) {
1022 case MLX4_STEERING_MODE_DEVICE_MANAGED:
1023 err = mlx4_flow_steer_promisc_add(mdev->dev,
1026 MLX4_FS_MC_DEFAULT);
1029 case MLX4_STEERING_MODE_B0:
1030 err = mlx4_multicast_promisc_add(mdev->dev,
1035 case MLX4_STEERING_MODE_A0:
1039 en_err(priv, "Failed entering multicast promisc mode\n");
1040 priv->flags |= MLX4_EN_FLAG_MC_PROMISC;
1043 /* Disable Multicast promisc */
1044 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1045 switch (mdev->dev->caps.steering_mode) {
1046 case MLX4_STEERING_MODE_DEVICE_MANAGED:
1047 err = mlx4_flow_steer_promisc_remove(mdev->dev,
1049 MLX4_FS_MC_DEFAULT);
1052 case MLX4_STEERING_MODE_B0:
1053 err = mlx4_multicast_promisc_remove(mdev->dev,
1058 case MLX4_STEERING_MODE_A0:
1062 en_err(priv, "Failed disabling multicast promiscuous mode\n");
1063 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1066 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
1067 0, MLX4_MCAST_DISABLE);
1069 en_err(priv, "Failed disabling multicast filter\n");
1071 /* Flush mcast filter and init it with broadcast address */
1072 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, ETH_BCAST,
1073 1, MLX4_MCAST_CONFIG);
1075 /* Update multicast list - we cache all addresses so they won't
1076 * change while HW is updated holding the command semaphor */
1077 netif_addr_lock_bh(dev);
1078 mlx4_en_cache_mclist(dev);
1079 netif_addr_unlock_bh(dev);
1080 list_for_each_entry(mclist, &priv->mc_list, list) {
1081 mcast_addr = mlx4_mac_to_u64(mclist->addr);
1082 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port,
1083 mcast_addr, 0, MLX4_MCAST_CONFIG);
1085 err = mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0,
1086 0, MLX4_MCAST_ENABLE);
1088 en_err(priv, "Failed enabling multicast filter\n");
1090 update_mclist_flags(priv, &priv->curr_list, &priv->mc_list);
1091 list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1092 if (mclist->action == MCLIST_REM) {
1093 /* detach this address and delete from list */
1094 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1095 mc_list[5] = priv->port;
1096 err = mlx4_multicast_detach(mdev->dev,
1097 &priv->rss_map.indir_qp,
1102 en_err(priv, "Fail to detach multicast address\n");
1104 if (mclist->tunnel_reg_id) {
1105 err = mlx4_flow_detach(priv->mdev->dev, mclist->tunnel_reg_id);
1107 en_err(priv, "Failed to detach multicast address\n");
1110 /* remove from list */
1111 list_del(&mclist->list);
1113 } else if (mclist->action == MCLIST_ADD) {
1114 /* attach the address */
1115 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1116 /* needed for B0 steering support */
1117 mc_list[5] = priv->port;
1118 err = mlx4_multicast_attach(mdev->dev,
1119 &priv->rss_map.indir_qp,
1125 en_err(priv, "Fail to attach multicast address\n");
1127 err = mlx4_en_tunnel_steer_add(priv, &mc_list[10], priv->base_qpn,
1128 &mclist->tunnel_reg_id);
1130 en_err(priv, "Failed to attach multicast address\n");
1136 static void mlx4_en_do_uc_filter(struct mlx4_en_priv *priv,
1137 struct net_device *dev,
1138 struct mlx4_en_dev *mdev)
1140 struct netdev_hw_addr *ha;
1141 struct mlx4_mac_entry *entry;
1142 struct hlist_node *tmp;
1146 struct hlist_head *bucket;
1151 /* Note that we do not need to protect our mac_hash traversal with rcu,
1152 * since all modification code is protected by mdev->state_lock
1155 /* find what to remove */
1156 for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i) {
1157 bucket = &priv->mac_hash[i];
1158 hlist_for_each_entry_safe(entry, tmp, bucket, hlist) {
1160 netdev_for_each_uc_addr(ha, dev) {
1161 if (ether_addr_equal_64bits(entry->mac,
1168 /* MAC address of the port is not in uc list */
1169 if (ether_addr_equal_64bits(entry->mac, dev->dev_addr))
1173 mac = mlx4_mac_to_u64(entry->mac);
1174 mlx4_en_uc_steer_release(priv, entry->mac,
1177 mlx4_unregister_mac(mdev->dev, priv->port, mac);
1179 hlist_del_rcu(&entry->hlist);
1180 kfree_rcu(entry, rcu);
1181 en_dbg(DRV, priv, "Removed MAC %pM on port:%d\n",
1182 entry->mac, priv->port);
1188 /* if we didn't remove anything, there is no use in trying to add
1189 * again once we are in a forced promisc mode state
1191 if ((priv->flags & MLX4_EN_FLAG_FORCE_PROMISC) && 0 == removed)
1194 prev_flags = priv->flags;
1195 priv->flags &= ~MLX4_EN_FLAG_FORCE_PROMISC;
1197 /* find what to add */
1198 netdev_for_each_uc_addr(ha, dev) {
1200 bucket = &priv->mac_hash[ha->addr[MLX4_EN_MAC_HASH_IDX]];
1201 hlist_for_each_entry(entry, bucket, hlist) {
1202 if (ether_addr_equal_64bits(entry->mac, ha->addr)) {
1209 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1211 en_err(priv, "Failed adding MAC %pM on port:%d (out of memory)\n",
1212 ha->addr, priv->port);
1213 priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1216 mac = mlx4_mac_to_u64(ha->addr);
1217 memcpy(entry->mac, ha->addr, ETH_ALEN);
1218 err = mlx4_register_mac(mdev->dev, priv->port, mac);
1220 en_err(priv, "Failed registering MAC %pM on port %d: %d\n",
1221 ha->addr, priv->port, err);
1223 priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1226 err = mlx4_en_uc_steer_add(priv, ha->addr,
1230 en_err(priv, "Failed adding MAC %pM on port %d: %d\n",
1231 ha->addr, priv->port, err);
1232 mlx4_unregister_mac(mdev->dev, priv->port, mac);
1234 priv->flags |= MLX4_EN_FLAG_FORCE_PROMISC;
1237 unsigned int mac_hash;
1238 en_dbg(DRV, priv, "Added MAC %pM on port:%d\n",
1239 ha->addr, priv->port);
1240 mac_hash = ha->addr[MLX4_EN_MAC_HASH_IDX];
1241 bucket = &priv->mac_hash[mac_hash];
1242 hlist_add_head_rcu(&entry->hlist, bucket);
1247 if (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC) {
1248 en_warn(priv, "Forcing promiscuous mode on port:%d\n",
1250 } else if (prev_flags & MLX4_EN_FLAG_FORCE_PROMISC) {
1251 en_warn(priv, "Stop forcing promiscuous mode on port:%d\n",
1256 static void mlx4_en_do_set_rx_mode(struct work_struct *work)
1258 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1260 struct mlx4_en_dev *mdev = priv->mdev;
1261 struct net_device *dev = priv->dev;
1263 mutex_lock(&mdev->state_lock);
1264 if (!mdev->device_up) {
1265 en_dbg(HW, priv, "Card is not up, ignoring rx mode change.\n");
1268 if (!priv->port_up) {
1269 en_dbg(HW, priv, "Port is down, ignoring rx mode change.\n");
1273 if (!netif_carrier_ok(dev)) {
1274 if (!mlx4_en_QUERY_PORT(mdev, priv->port)) {
1275 if (priv->port_state.link_state) {
1276 priv->last_link_state = MLX4_DEV_EVENT_PORT_UP;
1277 netif_carrier_on(dev);
1278 en_dbg(LINK, priv, "Link Up\n");
1283 if (dev->priv_flags & IFF_UNICAST_FLT)
1284 mlx4_en_do_uc_filter(priv, dev, mdev);
1286 /* Promsicuous mode: disable all filters */
1287 if ((dev->flags & IFF_PROMISC) ||
1288 (priv->flags & MLX4_EN_FLAG_FORCE_PROMISC)) {
1289 mlx4_en_set_promisc_mode(priv, mdev);
1293 /* Not in promiscuous mode */
1294 if (priv->flags & MLX4_EN_FLAG_PROMISC)
1295 mlx4_en_clear_promisc_mode(priv, mdev);
1297 mlx4_en_do_multicast(priv, dev, mdev);
1299 mutex_unlock(&mdev->state_lock);
1302 #ifdef CONFIG_NET_POLL_CONTROLLER
1303 static void mlx4_en_netpoll(struct net_device *dev)
1305 struct mlx4_en_priv *priv = netdev_priv(dev);
1306 struct mlx4_en_cq *cq;
1309 for (i = 0; i < priv->rx_ring_num; i++) {
1310 cq = priv->rx_cq[i];
1311 napi_schedule(&cq->napi);
1316 static void mlx4_en_tx_timeout(struct net_device *dev)
1318 struct mlx4_en_priv *priv = netdev_priv(dev);
1319 struct mlx4_en_dev *mdev = priv->mdev;
1322 if (netif_msg_timer(priv))
1323 en_warn(priv, "Tx timeout called on port:%d\n", priv->port);
1325 for (i = 0; i < priv->tx_ring_num; i++) {
1326 if (!netif_tx_queue_stopped(netdev_get_tx_queue(dev, i)))
1328 en_warn(priv, "TX timeout on queue: %d, QP: 0x%x, CQ: 0x%x, Cons: 0x%x, Prod: 0x%x\n",
1329 i, priv->tx_ring[i]->qpn, priv->tx_ring[i]->cqn,
1330 priv->tx_ring[i]->cons, priv->tx_ring[i]->prod);
1333 priv->port_stats.tx_timeout++;
1334 en_dbg(DRV, priv, "Scheduling watchdog\n");
1335 queue_work(mdev->workqueue, &priv->watchdog_task);
1339 static struct net_device_stats *mlx4_en_get_stats(struct net_device *dev)
1341 struct mlx4_en_priv *priv = netdev_priv(dev);
1343 spin_lock_bh(&priv->stats_lock);
1344 memcpy(&priv->ret_stats, &priv->stats, sizeof(priv->stats));
1345 spin_unlock_bh(&priv->stats_lock);
1347 return &priv->ret_stats;
1350 static void mlx4_en_set_default_moderation(struct mlx4_en_priv *priv)
1352 struct mlx4_en_cq *cq;
1355 /* If we haven't received a specific coalescing setting
1356 * (module param), we set the moderation parameters as follows:
1357 * - moder_cnt is set to the number of mtu sized packets to
1358 * satisfy our coalescing target.
1359 * - moder_time is set to a fixed value.
1361 priv->rx_frames = MLX4_EN_RX_COAL_TARGET;
1362 priv->rx_usecs = MLX4_EN_RX_COAL_TIME;
1363 priv->tx_frames = MLX4_EN_TX_COAL_PKTS;
1364 priv->tx_usecs = MLX4_EN_TX_COAL_TIME;
1365 en_dbg(INTR, priv, "Default coalesing params for mtu:%d - rx_frames:%d rx_usecs:%d\n",
1366 priv->dev->mtu, priv->rx_frames, priv->rx_usecs);
1368 /* Setup cq moderation params */
1369 for (i = 0; i < priv->rx_ring_num; i++) {
1370 cq = priv->rx_cq[i];
1371 cq->moder_cnt = priv->rx_frames;
1372 cq->moder_time = priv->rx_usecs;
1373 priv->last_moder_time[i] = MLX4_EN_AUTO_CONF;
1374 priv->last_moder_packets[i] = 0;
1375 priv->last_moder_bytes[i] = 0;
1378 for (i = 0; i < priv->tx_ring_num; i++) {
1379 cq = priv->tx_cq[i];
1380 cq->moder_cnt = priv->tx_frames;
1381 cq->moder_time = priv->tx_usecs;
1384 /* Reset auto-moderation params */
1385 priv->pkt_rate_low = MLX4_EN_RX_RATE_LOW;
1386 priv->rx_usecs_low = MLX4_EN_RX_COAL_TIME_LOW;
1387 priv->pkt_rate_high = MLX4_EN_RX_RATE_HIGH;
1388 priv->rx_usecs_high = MLX4_EN_RX_COAL_TIME_HIGH;
1389 priv->sample_interval = MLX4_EN_SAMPLE_INTERVAL;
1390 priv->adaptive_rx_coal = 1;
1391 priv->last_moder_jiffies = 0;
1392 priv->last_moder_tx_packets = 0;
1395 static void mlx4_en_auto_moderation(struct mlx4_en_priv *priv)
1397 unsigned long period = (unsigned long) (jiffies - priv->last_moder_jiffies);
1398 struct mlx4_en_cq *cq;
1399 unsigned long packets;
1401 unsigned long avg_pkt_size;
1402 unsigned long rx_packets;
1403 unsigned long rx_bytes;
1404 unsigned long rx_pkt_diff;
1408 if (!priv->adaptive_rx_coal || period < priv->sample_interval * HZ)
1411 for (ring = 0; ring < priv->rx_ring_num; ring++) {
1412 spin_lock_bh(&priv->stats_lock);
1413 rx_packets = priv->rx_ring[ring]->packets;
1414 rx_bytes = priv->rx_ring[ring]->bytes;
1415 spin_unlock_bh(&priv->stats_lock);
1417 rx_pkt_diff = ((unsigned long) (rx_packets -
1418 priv->last_moder_packets[ring]));
1419 packets = rx_pkt_diff;
1420 rate = packets * HZ / period;
1421 avg_pkt_size = packets ? ((unsigned long) (rx_bytes -
1422 priv->last_moder_bytes[ring])) / packets : 0;
1424 /* Apply auto-moderation only when packet rate
1425 * exceeds a rate that it matters */
1426 if (rate > (MLX4_EN_RX_RATE_THRESH / priv->rx_ring_num) &&
1427 avg_pkt_size > MLX4_EN_AVG_PKT_SMALL) {
1428 if (rate < priv->pkt_rate_low)
1429 moder_time = priv->rx_usecs_low;
1430 else if (rate > priv->pkt_rate_high)
1431 moder_time = priv->rx_usecs_high;
1433 moder_time = (rate - priv->pkt_rate_low) *
1434 (priv->rx_usecs_high - priv->rx_usecs_low) /
1435 (priv->pkt_rate_high - priv->pkt_rate_low) +
1438 moder_time = priv->rx_usecs_low;
1441 if (moder_time != priv->last_moder_time[ring]) {
1442 priv->last_moder_time[ring] = moder_time;
1443 cq = priv->rx_cq[ring];
1444 cq->moder_time = moder_time;
1445 cq->moder_cnt = priv->rx_frames;
1446 err = mlx4_en_set_cq_moder(priv, cq);
1448 en_err(priv, "Failed modifying moderation for cq:%d\n",
1451 priv->last_moder_packets[ring] = rx_packets;
1452 priv->last_moder_bytes[ring] = rx_bytes;
1455 priv->last_moder_jiffies = jiffies;
1458 static void mlx4_en_do_get_stats(struct work_struct *work)
1460 struct delayed_work *delay = to_delayed_work(work);
1461 struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1463 struct mlx4_en_dev *mdev = priv->mdev;
1466 mutex_lock(&mdev->state_lock);
1467 if (mdev->device_up) {
1468 if (priv->port_up) {
1469 err = mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 0);
1471 en_dbg(HW, priv, "Could not update stats\n");
1473 mlx4_en_auto_moderation(priv);
1476 queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
1478 if (mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port]) {
1479 mlx4_en_do_set_mac(priv);
1480 mdev->mac_removed[MLX4_MAX_PORTS + 1 - priv->port] = 0;
1482 mutex_unlock(&mdev->state_lock);
1485 /* mlx4_en_service_task - Run service task for tasks that needed to be done
1488 static void mlx4_en_service_task(struct work_struct *work)
1490 struct delayed_work *delay = to_delayed_work(work);
1491 struct mlx4_en_priv *priv = container_of(delay, struct mlx4_en_priv,
1493 struct mlx4_en_dev *mdev = priv->mdev;
1495 mutex_lock(&mdev->state_lock);
1496 if (mdev->device_up) {
1497 if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
1498 mlx4_en_ptp_overflow_check(mdev);
1500 queue_delayed_work(mdev->workqueue, &priv->service_task,
1501 SERVICE_TASK_DELAY);
1503 mutex_unlock(&mdev->state_lock);
1506 static void mlx4_en_linkstate(struct work_struct *work)
1508 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1510 struct mlx4_en_dev *mdev = priv->mdev;
1511 int linkstate = priv->link_state;
1513 mutex_lock(&mdev->state_lock);
1514 /* If observable port state changed set carrier state and
1515 * report to system log */
1516 if (priv->last_link_state != linkstate) {
1517 if (linkstate == MLX4_DEV_EVENT_PORT_DOWN) {
1518 en_info(priv, "Link Down\n");
1519 netif_carrier_off(priv->dev);
1521 en_info(priv, "Link Up\n");
1522 netif_carrier_on(priv->dev);
1525 priv->last_link_state = linkstate;
1526 mutex_unlock(&mdev->state_lock);
1529 static int mlx4_en_init_affinity_hint(struct mlx4_en_priv *priv, int ring_idx)
1531 struct mlx4_en_rx_ring *ring = priv->rx_ring[ring_idx];
1532 int numa_node = priv->mdev->dev->numa_node;
1535 if (!zalloc_cpumask_var(&ring->affinity_mask, GFP_KERNEL))
1538 ret = cpumask_set_cpu_local_first(ring_idx, numa_node,
1539 ring->affinity_mask);
1541 free_cpumask_var(ring->affinity_mask);
1546 static void mlx4_en_free_affinity_hint(struct mlx4_en_priv *priv, int ring_idx)
1548 free_cpumask_var(priv->rx_ring[ring_idx]->affinity_mask);
1551 int mlx4_en_start_port(struct net_device *dev)
1553 struct mlx4_en_priv *priv = netdev_priv(dev);
1554 struct mlx4_en_dev *mdev = priv->mdev;
1555 struct mlx4_en_cq *cq;
1556 struct mlx4_en_tx_ring *tx_ring;
1562 u8 mc_list[16] = {0};
1564 if (priv->port_up) {
1565 en_dbg(DRV, priv, "start port called while port already up\n");
1569 INIT_LIST_HEAD(&priv->mc_list);
1570 INIT_LIST_HEAD(&priv->curr_list);
1571 INIT_LIST_HEAD(&priv->ethtool_list);
1572 memset(&priv->ethtool_rules[0], 0,
1573 sizeof(struct ethtool_flow_id) * MAX_NUM_OF_FS_RULES);
1575 /* Calculate Rx buf size */
1576 dev->mtu = min(dev->mtu, priv->max_mtu);
1577 mlx4_en_calc_rx_buf(dev);
1578 en_dbg(DRV, priv, "Rx buf size:%d\n", priv->rx_skb_size);
1580 /* Configure rx cq's and rings */
1581 err = mlx4_en_activate_rx_rings(priv);
1583 en_err(priv, "Failed to activate RX rings\n");
1586 for (i = 0; i < priv->rx_ring_num; i++) {
1587 cq = priv->rx_cq[i];
1589 mlx4_en_cq_init_lock(cq);
1591 err = mlx4_en_init_affinity_hint(priv, i);
1593 en_err(priv, "Failed preparing IRQ affinity hint\n");
1597 err = mlx4_en_activate_cq(priv, cq, i);
1599 en_err(priv, "Failed activating Rx CQ\n");
1600 mlx4_en_free_affinity_hint(priv, i);
1603 for (j = 0; j < cq->size; j++)
1604 cq->buf[j].owner_sr_opcode = MLX4_CQE_OWNER_MASK;
1605 err = mlx4_en_set_cq_moder(priv, cq);
1607 en_err(priv, "Failed setting cq moderation parameters\n");
1608 mlx4_en_deactivate_cq(priv, cq);
1609 mlx4_en_free_affinity_hint(priv, i);
1612 mlx4_en_arm_cq(priv, cq);
1613 priv->rx_ring[i]->cqn = cq->mcq.cqn;
1618 en_dbg(DRV, priv, "Getting qp number for port %d\n", priv->port);
1619 err = mlx4_en_get_qp(priv);
1621 en_err(priv, "Failed getting eth qp\n");
1624 mdev->mac_removed[priv->port] = 0;
1626 err = mlx4_en_config_rss_steer(priv);
1628 en_err(priv, "Failed configuring rss steering\n");
1632 err = mlx4_en_create_drop_qp(priv);
1636 /* Configure tx cq's and rings */
1637 for (i = 0; i < priv->tx_ring_num; i++) {
1639 cq = priv->tx_cq[i];
1640 err = mlx4_en_activate_cq(priv, cq, i);
1642 en_err(priv, "Failed allocating Tx CQ\n");
1645 err = mlx4_en_set_cq_moder(priv, cq);
1647 en_err(priv, "Failed setting cq moderation parameters\n");
1648 mlx4_en_deactivate_cq(priv, cq);
1651 en_dbg(DRV, priv, "Resetting index of collapsed CQ:%d to -1\n", i);
1652 cq->buf->wqe_index = cpu_to_be16(0xffff);
1654 /* Configure ring */
1655 tx_ring = priv->tx_ring[i];
1656 err = mlx4_en_activate_tx_ring(priv, tx_ring, cq->mcq.cqn,
1657 i / priv->num_tx_rings_p_up);
1659 en_err(priv, "Failed allocating Tx ring\n");
1660 mlx4_en_deactivate_cq(priv, cq);
1663 tx_ring->tx_queue = netdev_get_tx_queue(dev, i);
1665 /* Arm CQ for TX completions */
1666 mlx4_en_arm_cq(priv, cq);
1668 /* Set initial ownership of all Tx TXBBs to SW (1) */
1669 for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
1670 *((u32 *) (tx_ring->buf + j)) = 0xffffffff;
1674 /* Configure port */
1675 err = mlx4_SET_PORT_general(mdev->dev, priv->port,
1676 priv->rx_skb_size + ETH_FCS_LEN,
1677 priv->prof->tx_pause,
1679 priv->prof->rx_pause,
1680 priv->prof->rx_ppp);
1682 en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
1686 /* Set default qp number */
1687 err = mlx4_SET_PORT_qpn_calc(mdev->dev, priv->port, priv->base_qpn, 0);
1689 en_err(priv, "Failed setting default qp numbers\n");
1693 if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
1694 err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC, 1);
1696 en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
1703 en_dbg(HW, priv, "Initializing port\n");
1704 err = mlx4_INIT_PORT(mdev->dev, priv->port);
1706 en_err(priv, "Failed Initializing port\n");
1710 /* Attach rx QP to bradcast address */
1711 memset(&mc_list[10], 0xff, ETH_ALEN);
1712 mc_list[5] = priv->port; /* needed for B0 steering support */
1713 if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1714 priv->port, 0, MLX4_PROT_ETH,
1715 &priv->broadcast_id))
1716 mlx4_warn(mdev, "Failed Attaching Broadcast\n");
1718 /* Must redo promiscuous mode setup. */
1719 priv->flags &= ~(MLX4_EN_FLAG_PROMISC | MLX4_EN_FLAG_MC_PROMISC);
1721 /* Schedule multicast task to populate multicast list */
1722 queue_work(mdev->workqueue, &priv->rx_mode_task);
1724 mlx4_set_stats_bitmap(mdev->dev, &priv->stats_bitmap);
1726 #ifdef CONFIG_MLX4_EN_VXLAN
1727 if (priv->mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_VXLAN_OFFLOADS)
1728 vxlan_get_rx_port(dev);
1730 priv->port_up = true;
1731 netif_tx_start_all_queues(dev);
1732 netif_device_attach(dev);
1737 while (tx_index--) {
1738 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[tx_index]);
1739 mlx4_en_deactivate_cq(priv, priv->tx_cq[tx_index]);
1741 mlx4_en_destroy_drop_qp(priv);
1743 mlx4_en_release_rss_steer(priv);
1745 mlx4_en_put_qp(priv);
1747 while (rx_index--) {
1748 mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
1749 mlx4_en_free_affinity_hint(priv, i);
1751 for (i = 0; i < priv->rx_ring_num; i++)
1752 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1754 return err; /* need to close devices */
1758 void mlx4_en_stop_port(struct net_device *dev, int detach)
1760 struct mlx4_en_priv *priv = netdev_priv(dev);
1761 struct mlx4_en_dev *mdev = priv->mdev;
1762 struct mlx4_en_mc_list *mclist, *tmp;
1763 struct ethtool_flow_id *flow, *tmp_flow;
1765 u8 mc_list[16] = {0};
1767 if (!priv->port_up) {
1768 en_dbg(DRV, priv, "stop port called while port already down\n");
1773 mlx4_CLOSE_PORT(mdev->dev, priv->port);
1775 /* Synchronize with tx routine */
1776 netif_tx_lock_bh(dev);
1778 netif_device_detach(dev);
1779 netif_tx_stop_all_queues(dev);
1780 netif_tx_unlock_bh(dev);
1782 netif_tx_disable(dev);
1784 /* Set port as not active */
1785 priv->port_up = false;
1787 /* Promsicuous mode */
1788 if (mdev->dev->caps.steering_mode ==
1789 MLX4_STEERING_MODE_DEVICE_MANAGED) {
1790 priv->flags &= ~(MLX4_EN_FLAG_PROMISC |
1791 MLX4_EN_FLAG_MC_PROMISC);
1792 mlx4_flow_steer_promisc_remove(mdev->dev,
1794 MLX4_FS_ALL_DEFAULT);
1795 mlx4_flow_steer_promisc_remove(mdev->dev,
1797 MLX4_FS_MC_DEFAULT);
1798 } else if (priv->flags & MLX4_EN_FLAG_PROMISC) {
1799 priv->flags &= ~MLX4_EN_FLAG_PROMISC;
1801 /* Disable promiscouos mode */
1802 mlx4_unicast_promisc_remove(mdev->dev, priv->base_qpn,
1805 /* Disable Multicast promisc */
1806 if (priv->flags & MLX4_EN_FLAG_MC_PROMISC) {
1807 mlx4_multicast_promisc_remove(mdev->dev, priv->base_qpn,
1809 priv->flags &= ~MLX4_EN_FLAG_MC_PROMISC;
1813 /* Detach All multicasts */
1814 memset(&mc_list[10], 0xff, ETH_ALEN);
1815 mc_list[5] = priv->port; /* needed for B0 steering support */
1816 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list,
1817 MLX4_PROT_ETH, priv->broadcast_id);
1818 list_for_each_entry(mclist, &priv->curr_list, list) {
1819 memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
1820 mc_list[5] = priv->port;
1821 mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp,
1822 mc_list, MLX4_PROT_ETH, mclist->reg_id);
1823 if (mclist->tunnel_reg_id)
1824 mlx4_flow_detach(mdev->dev, mclist->tunnel_reg_id);
1826 mlx4_en_clear_list(dev);
1827 list_for_each_entry_safe(mclist, tmp, &priv->curr_list, list) {
1828 list_del(&mclist->list);
1832 /* Flush multicast filter */
1833 mlx4_SET_MCAST_FLTR(mdev->dev, priv->port, 0, 1, MLX4_MCAST_CONFIG);
1835 /* Remove flow steering rules for the port*/
1836 if (mdev->dev->caps.steering_mode ==
1837 MLX4_STEERING_MODE_DEVICE_MANAGED) {
1839 list_for_each_entry_safe(flow, tmp_flow,
1840 &priv->ethtool_list, list) {
1841 mlx4_flow_detach(mdev->dev, flow->id);
1842 list_del(&flow->list);
1846 mlx4_en_destroy_drop_qp(priv);
1849 for (i = 0; i < priv->tx_ring_num; i++) {
1850 mlx4_en_deactivate_tx_ring(priv, priv->tx_ring[i]);
1851 mlx4_en_deactivate_cq(priv, priv->tx_cq[i]);
1855 for (i = 0; i < priv->tx_ring_num; i++)
1856 mlx4_en_free_tx_buf(dev, priv->tx_ring[i]);
1859 mlx4_en_release_rss_steer(priv);
1861 /* Unregister Mac address for the port */
1862 mlx4_en_put_qp(priv);
1863 if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_REASSIGN_MAC_EN))
1864 mdev->mac_removed[priv->port] = 1;
1867 for (i = 0; i < priv->rx_ring_num; i++) {
1868 struct mlx4_en_cq *cq = priv->rx_cq[i];
1871 while (!mlx4_en_cq_lock_napi(cq)) {
1872 pr_info("CQ %d locked\n", i);
1877 while (test_bit(NAPI_STATE_SCHED, &cq->napi.state))
1879 mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
1880 mlx4_en_deactivate_cq(priv, cq);
1882 mlx4_en_free_affinity_hint(priv, i);
1886 static void mlx4_en_restart(struct work_struct *work)
1888 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
1890 struct mlx4_en_dev *mdev = priv->mdev;
1891 struct net_device *dev = priv->dev;
1893 en_dbg(DRV, priv, "Watchdog task called for port %d\n", priv->port);
1895 mutex_lock(&mdev->state_lock);
1896 if (priv->port_up) {
1897 mlx4_en_stop_port(dev, 1);
1898 if (mlx4_en_start_port(dev))
1899 en_err(priv, "Failed restarting port %d\n", priv->port);
1901 mutex_unlock(&mdev->state_lock);
1904 static void mlx4_en_clear_stats(struct net_device *dev)
1906 struct mlx4_en_priv *priv = netdev_priv(dev);
1907 struct mlx4_en_dev *mdev = priv->mdev;
1910 if (mlx4_en_DUMP_ETH_STATS(mdev, priv->port, 1))
1911 en_dbg(HW, priv, "Failed dumping statistics\n");
1913 memset(&priv->stats, 0, sizeof(priv->stats));
1914 memset(&priv->pstats, 0, sizeof(priv->pstats));
1915 memset(&priv->pkstats, 0, sizeof(priv->pkstats));
1916 memset(&priv->port_stats, 0, sizeof(priv->port_stats));
1918 for (i = 0; i < priv->tx_ring_num; i++) {
1919 priv->tx_ring[i]->bytes = 0;
1920 priv->tx_ring[i]->packets = 0;
1921 priv->tx_ring[i]->tx_csum = 0;
1923 for (i = 0; i < priv->rx_ring_num; i++) {
1924 priv->rx_ring[i]->bytes = 0;
1925 priv->rx_ring[i]->packets = 0;
1926 priv->rx_ring[i]->csum_ok = 0;
1927 priv->rx_ring[i]->csum_none = 0;
1931 static int mlx4_en_open(struct net_device *dev)
1933 struct mlx4_en_priv *priv = netdev_priv(dev);
1934 struct mlx4_en_dev *mdev = priv->mdev;
1937 mutex_lock(&mdev->state_lock);
1939 if (!mdev->device_up) {
1940 en_err(priv, "Cannot open - device down/disabled\n");
1945 /* Reset HW statistics and SW counters */
1946 mlx4_en_clear_stats(dev);
1948 err = mlx4_en_start_port(dev);
1950 en_err(priv, "Failed starting port:%d\n", priv->port);
1953 mutex_unlock(&mdev->state_lock);
1958 static int mlx4_en_close(struct net_device *dev)
1960 struct mlx4_en_priv *priv = netdev_priv(dev);
1961 struct mlx4_en_dev *mdev = priv->mdev;
1963 en_dbg(IFDOWN, priv, "Close port called\n");
1965 mutex_lock(&mdev->state_lock);
1967 mlx4_en_stop_port(dev, 0);
1968 netif_carrier_off(dev);
1970 mutex_unlock(&mdev->state_lock);
1974 void mlx4_en_free_resources(struct mlx4_en_priv *priv)
1978 #ifdef CONFIG_RFS_ACCEL
1979 free_irq_cpu_rmap(priv->dev->rx_cpu_rmap);
1980 priv->dev->rx_cpu_rmap = NULL;
1983 for (i = 0; i < priv->tx_ring_num; i++) {
1984 if (priv->tx_ring && priv->tx_ring[i])
1985 mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
1986 if (priv->tx_cq && priv->tx_cq[i])
1987 mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
1990 for (i = 0; i < priv->rx_ring_num; i++) {
1991 if (priv->rx_ring[i])
1992 mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
1993 priv->prof->rx_ring_size, priv->stride);
1995 mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
1998 if (priv->base_tx_qpn) {
1999 mlx4_qp_release_range(priv->mdev->dev, priv->base_tx_qpn, priv->tx_ring_num);
2000 priv->base_tx_qpn = 0;
2004 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv)
2006 struct mlx4_en_port_profile *prof = priv->prof;
2011 err = mlx4_qp_reserve_range(priv->mdev->dev, priv->tx_ring_num, 256, &priv->base_tx_qpn);
2013 en_err(priv, "failed reserving range for TX rings\n");
2017 /* Create tx Rings */
2018 for (i = 0; i < priv->tx_ring_num; i++) {
2019 node = cpu_to_node(i % num_online_cpus());
2020 if (mlx4_en_create_cq(priv, &priv->tx_cq[i],
2021 prof->tx_ring_size, i, TX, node))
2024 if (mlx4_en_create_tx_ring(priv, &priv->tx_ring[i],
2025 priv->base_tx_qpn + i,
2026 prof->tx_ring_size, TXBB_SIZE,
2031 /* Create rx Rings */
2032 for (i = 0; i < priv->rx_ring_num; i++) {
2033 node = cpu_to_node(i % num_online_cpus());
2034 if (mlx4_en_create_cq(priv, &priv->rx_cq[i],
2035 prof->rx_ring_size, i, RX, node))
2038 if (mlx4_en_create_rx_ring(priv, &priv->rx_ring[i],
2039 prof->rx_ring_size, priv->stride,
2044 #ifdef CONFIG_RFS_ACCEL
2045 if (priv->mdev->dev->caps.comp_pool) {
2046 priv->dev->rx_cpu_rmap = alloc_irq_cpu_rmap(priv->mdev->dev->caps.comp_pool);
2047 if (!priv->dev->rx_cpu_rmap)
2055 en_err(priv, "Failed to allocate NIC resources\n");
2056 for (i = 0; i < priv->rx_ring_num; i++) {
2057 if (priv->rx_ring[i])
2058 mlx4_en_destroy_rx_ring(priv, &priv->rx_ring[i],
2062 mlx4_en_destroy_cq(priv, &priv->rx_cq[i]);
2064 for (i = 0; i < priv->tx_ring_num; i++) {
2065 if (priv->tx_ring[i])
2066 mlx4_en_destroy_tx_ring(priv, &priv->tx_ring[i]);
2068 mlx4_en_destroy_cq(priv, &priv->tx_cq[i]);
2074 void mlx4_en_destroy_netdev(struct net_device *dev)
2076 struct mlx4_en_priv *priv = netdev_priv(dev);
2077 struct mlx4_en_dev *mdev = priv->mdev;
2079 en_dbg(DRV, priv, "Destroying netdev on port:%d\n", priv->port);
2081 /* Unregister device - this will close the port if it was up */
2082 if (priv->registered)
2083 unregister_netdev(dev);
2085 if (priv->allocated)
2086 mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
2088 cancel_delayed_work(&priv->stats_task);
2089 cancel_delayed_work(&priv->service_task);
2090 /* flush any pending task for this netdev */
2091 flush_workqueue(mdev->workqueue);
2093 /* Detach the netdev so tasks would not attempt to access it */
2094 mutex_lock(&mdev->state_lock);
2095 mdev->pndev[priv->port] = NULL;
2096 mutex_unlock(&mdev->state_lock);
2098 mlx4_en_free_resources(priv);
2100 kfree(priv->tx_ring);
2106 static int mlx4_en_change_mtu(struct net_device *dev, int new_mtu)
2108 struct mlx4_en_priv *priv = netdev_priv(dev);
2109 struct mlx4_en_dev *mdev = priv->mdev;
2112 en_dbg(DRV, priv, "Change MTU called - current:%d new:%d\n",
2115 if ((new_mtu < MLX4_EN_MIN_MTU) || (new_mtu > priv->max_mtu)) {
2116 en_err(priv, "Bad MTU size:%d.\n", new_mtu);
2121 if (netif_running(dev)) {
2122 mutex_lock(&mdev->state_lock);
2123 if (!mdev->device_up) {
2124 /* NIC is probably restarting - let watchdog task reset
2126 en_dbg(DRV, priv, "Change MTU called with card down!?\n");
2128 mlx4_en_stop_port(dev, 1);
2129 err = mlx4_en_start_port(dev);
2131 en_err(priv, "Failed restarting port:%d\n",
2133 queue_work(mdev->workqueue, &priv->watchdog_task);
2136 mutex_unlock(&mdev->state_lock);
2141 static int mlx4_en_hwtstamp_set(struct net_device *dev, struct ifreq *ifr)
2143 struct mlx4_en_priv *priv = netdev_priv(dev);
2144 struct mlx4_en_dev *mdev = priv->mdev;
2145 struct hwtstamp_config config;
2147 if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
2150 /* reserved for future extensions */
2154 /* device doesn't support time stamping */
2155 if (!(mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS))
2158 /* TX HW timestamp */
2159 switch (config.tx_type) {
2160 case HWTSTAMP_TX_OFF:
2161 case HWTSTAMP_TX_ON:
2167 /* RX HW timestamp */
2168 switch (config.rx_filter) {
2169 case HWTSTAMP_FILTER_NONE:
2171 case HWTSTAMP_FILTER_ALL:
2172 case HWTSTAMP_FILTER_SOME:
2173 case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
2174 case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
2175 case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
2176 case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
2177 case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
2178 case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
2179 case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
2180 case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
2181 case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
2182 case HWTSTAMP_FILTER_PTP_V2_EVENT:
2183 case HWTSTAMP_FILTER_PTP_V2_SYNC:
2184 case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
2185 config.rx_filter = HWTSTAMP_FILTER_ALL;
2191 if (mlx4_en_timestamp_config(dev, config.tx_type, config.rx_filter)) {
2192 config.tx_type = HWTSTAMP_TX_OFF;
2193 config.rx_filter = HWTSTAMP_FILTER_NONE;
2196 return copy_to_user(ifr->ifr_data, &config,
2197 sizeof(config)) ? -EFAULT : 0;
2200 static int mlx4_en_hwtstamp_get(struct net_device *dev, struct ifreq *ifr)
2202 struct mlx4_en_priv *priv = netdev_priv(dev);
2204 return copy_to_user(ifr->ifr_data, &priv->hwtstamp_config,
2205 sizeof(priv->hwtstamp_config)) ? -EFAULT : 0;
2208 static int mlx4_en_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
2212 return mlx4_en_hwtstamp_set(dev, ifr);
2214 return mlx4_en_hwtstamp_get(dev, ifr);
2220 static int mlx4_en_set_features(struct net_device *netdev,
2221 netdev_features_t features)
2223 struct mlx4_en_priv *priv = netdev_priv(netdev);
2225 if (features & NETIF_F_LOOPBACK)
2226 priv->ctrl_flags |= cpu_to_be32(MLX4_WQE_CTRL_FORCE_LOOPBACK);
2229 cpu_to_be32(~MLX4_WQE_CTRL_FORCE_LOOPBACK);
2231 mlx4_en_update_loopback_state(netdev, features);
2237 static int mlx4_en_set_vf_mac(struct net_device *dev, int queue, u8 *mac)
2239 struct mlx4_en_priv *en_priv = netdev_priv(dev);
2240 struct mlx4_en_dev *mdev = en_priv->mdev;
2241 u64 mac_u64 = mlx4_mac_to_u64(mac);
2243 if (!is_valid_ether_addr(mac))
2246 return mlx4_set_vf_mac(mdev->dev, en_priv->port, queue, mac_u64);
2249 static int mlx4_en_set_vf_vlan(struct net_device *dev, int vf, u16 vlan, u8 qos)
2251 struct mlx4_en_priv *en_priv = netdev_priv(dev);
2252 struct mlx4_en_dev *mdev = en_priv->mdev;
2254 return mlx4_set_vf_vlan(mdev->dev, en_priv->port, vf, vlan, qos);
2257 static int mlx4_en_set_vf_spoofchk(struct net_device *dev, int vf, bool setting)
2259 struct mlx4_en_priv *en_priv = netdev_priv(dev);
2260 struct mlx4_en_dev *mdev = en_priv->mdev;
2262 return mlx4_set_vf_spoofchk(mdev->dev, en_priv->port, vf, setting);
2265 static int mlx4_en_get_vf_config(struct net_device *dev, int vf, struct ifla_vf_info *ivf)
2267 struct mlx4_en_priv *en_priv = netdev_priv(dev);
2268 struct mlx4_en_dev *mdev = en_priv->mdev;
2270 return mlx4_get_vf_config(mdev->dev, en_priv->port, vf, ivf);
2273 static int mlx4_en_set_vf_link_state(struct net_device *dev, int vf, int link_state)
2275 struct mlx4_en_priv *en_priv = netdev_priv(dev);
2276 struct mlx4_en_dev *mdev = en_priv->mdev;
2278 return mlx4_set_vf_link_state(mdev->dev, en_priv->port, vf, link_state);
2281 #define PORT_ID_BYTE_LEN 8
2282 static int mlx4_en_get_phys_port_id(struct net_device *dev,
2283 struct netdev_phys_port_id *ppid)
2285 struct mlx4_en_priv *priv = netdev_priv(dev);
2286 struct mlx4_dev *mdev = priv->mdev->dev;
2288 u64 phys_port_id = mdev->caps.phys_port_id[priv->port];
2293 ppid->id_len = sizeof(phys_port_id);
2294 for (i = PORT_ID_BYTE_LEN - 1; i >= 0; --i) {
2295 ppid->id[i] = phys_port_id & 0xff;
2301 #ifdef CONFIG_MLX4_EN_VXLAN
2302 static void mlx4_en_add_vxlan_offloads(struct work_struct *work)
2305 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2308 ret = mlx4_config_vxlan_port(priv->mdev->dev, priv->vxlan_port);
2312 ret = mlx4_SET_PORT_VXLAN(priv->mdev->dev, priv->port,
2313 VXLAN_STEER_BY_OUTER_MAC, 1);
2316 en_err(priv, "failed setting L2 tunnel configuration ret %d\n", ret);
2319 static void mlx4_en_del_vxlan_offloads(struct work_struct *work)
2322 struct mlx4_en_priv *priv = container_of(work, struct mlx4_en_priv,
2325 ret = mlx4_SET_PORT_VXLAN(priv->mdev->dev, priv->port,
2326 VXLAN_STEER_BY_OUTER_MAC, 0);
2328 en_err(priv, "failed setting L2 tunnel configuration ret %d\n", ret);
2330 priv->vxlan_port = 0;
2333 static void mlx4_en_add_vxlan_port(struct net_device *dev,
2334 sa_family_t sa_family, __be16 port)
2336 struct mlx4_en_priv *priv = netdev_priv(dev);
2337 __be16 current_port;
2339 if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
2342 if (sa_family == AF_INET6)
2345 current_port = priv->vxlan_port;
2346 if (current_port && current_port != port) {
2347 en_warn(priv, "vxlan port %d configured, can't add port %d\n",
2348 ntohs(current_port), ntohs(port));
2352 priv->vxlan_port = port;
2353 queue_work(priv->mdev->workqueue, &priv->vxlan_add_task);
2356 static void mlx4_en_del_vxlan_port(struct net_device *dev,
2357 sa_family_t sa_family, __be16 port)
2359 struct mlx4_en_priv *priv = netdev_priv(dev);
2360 __be16 current_port;
2362 if (priv->mdev->dev->caps.tunnel_offload_mode != MLX4_TUNNEL_OFFLOAD_MODE_VXLAN)
2365 if (sa_family == AF_INET6)
2368 current_port = priv->vxlan_port;
2369 if (current_port != port) {
2370 en_dbg(DRV, priv, "vxlan port %d isn't configured, ignoring\n", ntohs(port));
2374 queue_work(priv->mdev->workqueue, &priv->vxlan_del_task);
2378 static const struct net_device_ops mlx4_netdev_ops = {
2379 .ndo_open = mlx4_en_open,
2380 .ndo_stop = mlx4_en_close,
2381 .ndo_start_xmit = mlx4_en_xmit,
2382 .ndo_select_queue = mlx4_en_select_queue,
2383 .ndo_get_stats = mlx4_en_get_stats,
2384 .ndo_set_rx_mode = mlx4_en_set_rx_mode,
2385 .ndo_set_mac_address = mlx4_en_set_mac,
2386 .ndo_validate_addr = eth_validate_addr,
2387 .ndo_change_mtu = mlx4_en_change_mtu,
2388 .ndo_do_ioctl = mlx4_en_ioctl,
2389 .ndo_tx_timeout = mlx4_en_tx_timeout,
2390 .ndo_vlan_rx_add_vid = mlx4_en_vlan_rx_add_vid,
2391 .ndo_vlan_rx_kill_vid = mlx4_en_vlan_rx_kill_vid,
2392 #ifdef CONFIG_NET_POLL_CONTROLLER
2393 .ndo_poll_controller = mlx4_en_netpoll,
2395 .ndo_set_features = mlx4_en_set_features,
2396 .ndo_setup_tc = mlx4_en_setup_tc,
2397 #ifdef CONFIG_RFS_ACCEL
2398 .ndo_rx_flow_steer = mlx4_en_filter_rfs,
2400 #ifdef CONFIG_NET_RX_BUSY_POLL
2401 .ndo_busy_poll = mlx4_en_low_latency_recv,
2403 .ndo_get_phys_port_id = mlx4_en_get_phys_port_id,
2404 #ifdef CONFIG_MLX4_EN_VXLAN
2405 .ndo_add_vxlan_port = mlx4_en_add_vxlan_port,
2406 .ndo_del_vxlan_port = mlx4_en_del_vxlan_port,
2410 static const struct net_device_ops mlx4_netdev_ops_master = {
2411 .ndo_open = mlx4_en_open,
2412 .ndo_stop = mlx4_en_close,
2413 .ndo_start_xmit = mlx4_en_xmit,
2414 .ndo_select_queue = mlx4_en_select_queue,
2415 .ndo_get_stats = mlx4_en_get_stats,
2416 .ndo_set_rx_mode = mlx4_en_set_rx_mode,
2417 .ndo_set_mac_address = mlx4_en_set_mac,
2418 .ndo_validate_addr = eth_validate_addr,
2419 .ndo_change_mtu = mlx4_en_change_mtu,
2420 .ndo_tx_timeout = mlx4_en_tx_timeout,
2421 .ndo_vlan_rx_add_vid = mlx4_en_vlan_rx_add_vid,
2422 .ndo_vlan_rx_kill_vid = mlx4_en_vlan_rx_kill_vid,
2423 .ndo_set_vf_mac = mlx4_en_set_vf_mac,
2424 .ndo_set_vf_vlan = mlx4_en_set_vf_vlan,
2425 .ndo_set_vf_spoofchk = mlx4_en_set_vf_spoofchk,
2426 .ndo_set_vf_link_state = mlx4_en_set_vf_link_state,
2427 .ndo_get_vf_config = mlx4_en_get_vf_config,
2428 #ifdef CONFIG_NET_POLL_CONTROLLER
2429 .ndo_poll_controller = mlx4_en_netpoll,
2431 .ndo_set_features = mlx4_en_set_features,
2432 .ndo_setup_tc = mlx4_en_setup_tc,
2433 #ifdef CONFIG_RFS_ACCEL
2434 .ndo_rx_flow_steer = mlx4_en_filter_rfs,
2436 .ndo_get_phys_port_id = mlx4_en_get_phys_port_id,
2439 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
2440 struct mlx4_en_port_profile *prof)
2442 struct net_device *dev;
2443 struct mlx4_en_priv *priv;
2448 dev = alloc_etherdev_mqs(sizeof(struct mlx4_en_priv),
2449 MAX_TX_RINGS, MAX_RX_RINGS);
2453 netif_set_real_num_tx_queues(dev, prof->tx_ring_num);
2454 netif_set_real_num_rx_queues(dev, prof->rx_ring_num);
2456 SET_NETDEV_DEV(dev, &mdev->dev->pdev->dev);
2457 dev->dev_port = port - 1;
2460 * Initialize driver private data
2463 priv = netdev_priv(dev);
2464 memset(priv, 0, sizeof(struct mlx4_en_priv));
2467 priv->ddev = &mdev->pdev->dev;
2470 priv->port_up = false;
2471 priv->flags = prof->flags;
2472 priv->ctrl_flags = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE |
2473 MLX4_WQE_CTRL_SOLICITED);
2474 priv->num_tx_rings_p_up = mdev->profile.num_tx_rings_p_up;
2475 priv->tx_ring_num = prof->tx_ring_num;
2476 priv->tx_work_limit = MLX4_EN_DEFAULT_TX_WORK;
2478 priv->tx_ring = kzalloc(sizeof(struct mlx4_en_tx_ring *) * MAX_TX_RINGS,
2480 if (!priv->tx_ring) {
2484 priv->tx_cq = kzalloc(sizeof(struct mlx4_en_cq *) * MAX_TX_RINGS,
2490 priv->rx_ring_num = prof->rx_ring_num;
2491 priv->cqe_factor = (mdev->dev->caps.cqe_size == 64) ? 1 : 0;
2492 priv->mac_index = -1;
2493 priv->msg_enable = MLX4_EN_MSG_LEVEL;
2494 spin_lock_init(&priv->stats_lock);
2495 INIT_WORK(&priv->rx_mode_task, mlx4_en_do_set_rx_mode);
2496 INIT_WORK(&priv->watchdog_task, mlx4_en_restart);
2497 INIT_WORK(&priv->linkstate_task, mlx4_en_linkstate);
2498 INIT_DELAYED_WORK(&priv->stats_task, mlx4_en_do_get_stats);
2499 INIT_DELAYED_WORK(&priv->service_task, mlx4_en_service_task);
2500 #ifdef CONFIG_MLX4_EN_VXLAN
2501 INIT_WORK(&priv->vxlan_add_task, mlx4_en_add_vxlan_offloads);
2502 INIT_WORK(&priv->vxlan_del_task, mlx4_en_del_vxlan_offloads);
2504 #ifdef CONFIG_MLX4_EN_DCB
2505 if (!mlx4_is_slave(priv->mdev->dev)) {
2506 if (mdev->dev->caps.flags & MLX4_DEV_CAP_FLAG_SET_ETH_SCHED) {
2507 dev->dcbnl_ops = &mlx4_en_dcbnl_ops;
2509 en_info(priv, "enabling only PFC DCB ops\n");
2510 dev->dcbnl_ops = &mlx4_en_dcbnl_pfc_ops;
2515 for (i = 0; i < MLX4_EN_MAC_HASH_SIZE; ++i)
2516 INIT_HLIST_HEAD(&priv->mac_hash[i]);
2518 /* Query for default mac and max mtu */
2519 priv->max_mtu = mdev->dev->caps.eth_mtu_cap[priv->port];
2521 /* Set default MAC */
2522 dev->addr_len = ETH_ALEN;
2523 mlx4_en_u64_to_mac(dev->dev_addr, mdev->dev->caps.def_mac[priv->port]);
2524 if (!is_valid_ether_addr(dev->dev_addr)) {
2525 if (mlx4_is_slave(priv->mdev->dev)) {
2526 eth_hw_addr_random(dev);
2527 en_warn(priv, "Assigned random MAC address %pM\n", dev->dev_addr);
2528 mac_u64 = mlx4_mac_to_u64(dev->dev_addr);
2529 mdev->dev->caps.def_mac[priv->port] = mac_u64;
2531 en_err(priv, "Port: %d, invalid mac burned: %pM, quiting\n",
2532 priv->port, dev->dev_addr);
2538 memcpy(priv->prev_mac, dev->dev_addr, sizeof(priv->prev_mac));
2540 priv->stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
2541 DS_SIZE * MLX4_EN_MAX_RX_FRAGS);
2542 err = mlx4_en_alloc_resources(priv);
2546 #ifdef CONFIG_RFS_ACCEL
2547 INIT_LIST_HEAD(&priv->filters);
2548 spin_lock_init(&priv->filters_lock);
2551 /* Initialize time stamping config */
2552 priv->hwtstamp_config.flags = 0;
2553 priv->hwtstamp_config.tx_type = HWTSTAMP_TX_OFF;
2554 priv->hwtstamp_config.rx_filter = HWTSTAMP_FILTER_NONE;
2556 /* Allocate page for receive rings */
2557 err = mlx4_alloc_hwq_res(mdev->dev, &priv->res,
2558 MLX4_EN_PAGE_SIZE, MLX4_EN_PAGE_SIZE);
2560 en_err(priv, "Failed to allocate page for rx qps\n");
2563 priv->allocated = 1;
2566 * Initialize netdev entry points
2568 if (mlx4_is_master(priv->mdev->dev))
2569 dev->netdev_ops = &mlx4_netdev_ops_master;
2571 dev->netdev_ops = &mlx4_netdev_ops;
2572 dev->watchdog_timeo = MLX4_EN_WATCHDOG_TIMEOUT;
2573 netif_set_real_num_tx_queues(dev, priv->tx_ring_num);
2574 netif_set_real_num_rx_queues(dev, priv->rx_ring_num);
2576 dev->ethtool_ops = &mlx4_en_ethtool_ops;
2579 * Set driver features
2581 dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
2582 if (mdev->LSO_support)
2583 dev->hw_features |= NETIF_F_TSO | NETIF_F_TSO6;
2585 dev->vlan_features = dev->hw_features;
2587 dev->hw_features |= NETIF_F_RXCSUM | NETIF_F_RXHASH;
2588 dev->features = dev->hw_features | NETIF_F_HIGHDMA |
2589 NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX |
2590 NETIF_F_HW_VLAN_CTAG_FILTER;
2591 dev->hw_features |= NETIF_F_LOOPBACK;
2593 if (mdev->dev->caps.steering_mode ==
2594 MLX4_STEERING_MODE_DEVICE_MANAGED)
2595 dev->hw_features |= NETIF_F_NTUPLE;
2597 if (mdev->dev->caps.steering_mode != MLX4_STEERING_MODE_A0)
2598 dev->priv_flags |= IFF_UNICAST_FLT;
2600 if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
2601 dev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_RXCSUM |
2602 NETIF_F_TSO | NETIF_F_GSO_UDP_TUNNEL;
2603 dev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
2604 dev->features |= NETIF_F_GSO_UDP_TUNNEL;
2607 mdev->pndev[port] = dev;
2609 netif_carrier_off(dev);
2610 mlx4_en_set_default_moderation(priv);
2612 err = register_netdev(dev);
2614 en_err(priv, "Netdev registration failed for port %d\n", port);
2617 priv->registered = 1;
2619 en_warn(priv, "Using %d TX rings\n", prof->tx_ring_num);
2620 en_warn(priv, "Using %d RX rings\n", prof->rx_ring_num);
2622 mlx4_en_update_loopback_state(priv->dev, priv->dev->features);
2624 /* Configure port */
2625 mlx4_en_calc_rx_buf(dev);
2626 err = mlx4_SET_PORT_general(mdev->dev, priv->port,
2627 priv->rx_skb_size + ETH_FCS_LEN,
2628 prof->tx_pause, prof->tx_ppp,
2629 prof->rx_pause, prof->rx_ppp);
2631 en_err(priv, "Failed setting port general configurations for port %d, with error %d\n",
2636 if (mdev->dev->caps.tunnel_offload_mode == MLX4_TUNNEL_OFFLOAD_MODE_VXLAN) {
2637 err = mlx4_SET_PORT_VXLAN(mdev->dev, priv->port, VXLAN_STEER_BY_OUTER_MAC, 1);
2639 en_err(priv, "Failed setting port L2 tunnel configuration, err %d\n",
2646 en_warn(priv, "Initializing port\n");
2647 err = mlx4_INIT_PORT(mdev->dev, priv->port);
2649 en_err(priv, "Failed Initializing port\n");
2652 queue_delayed_work(mdev->workqueue, &priv->stats_task, STATS_DELAY);
2654 if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_TS)
2655 queue_delayed_work(mdev->workqueue, &priv->service_task,
2656 SERVICE_TASK_DELAY);
2661 mlx4_en_destroy_netdev(dev);